You are on page 1of 6

PROXY MEANS TESTING: AN ALTERNATIVE METHOD FOR

POVERTY ASSESSMENT
SURAPONE PTANAWANIT

Faculty of Social Administration, Thammasat University, Bangkok, Thailand


E-mail: suraponep@hotmail.com

Abstract- This article discusses the experiment of a proxy means test for assessing poverty levels in Thailand. It explains
why a proxy means test instead of a conventional income or assets test is more applicable to a developing country like
Thailand where income data are rather inaccurate. The main point of the article is its discussion in details on how the study
develops the proxy means test. Apart from suggesting a way by which poverty can be measured by the proxy means test, the
article also explains how this instrument can assess clients’ eligibility for social assistance benefit.

Keywords- Poverty Assessment, Eligibility Assessment, Proxy Means Test, Social Assistance.

I. INTRODUCTION another problem with which we are most familiar.


Stigma discourages the target people to exercise their
Direct poverty assessment such as a means test and an rights and results in a waste of welfare resources due to
assets test has been a common tool for evaluating the a low take-up rate of means tested benefit (see Moffitt
right to social assistance benefit in welfare states. 1983; Pudney, Hancock, & Sutherland 2006)
Although this direct poverty assessment has become a In developing countries, limited application of means
useful instrument for social assistance arrangements, its tests is not because of its complexity and cause of
applications in many occasions bring about negative stigmatization; but unreliable income data and
consequences. As direct poverty testing is concerned inaccurate poverty lines are the main reason. Kidd &
with financial status, eligible clients are labeled as Wyld (2011) argue that targeting the poor, especially, in
people with poverty. This labeling problem is developing countries is complicate because their large
sometimes believed to be the main cause of a low take- informal economies make it difficult to accurately assess
up rate. van Oorschot and Schell (1989) indicate in their people’s incomes. Other authors including Johannsen
study that the decline in the social assistance take-up rate (2006) mention similar limitations of using proxy means
might be the result of social labeling in Western tests for targeting the poor and for providing welfare in
European countries such as the UK, the Netherlands, these countries. Instead of encouraging use of means
and Germany. Their study also showed that the non- testing, some international organizations tend to support
take-up rate for the family income supplement in the UK application of a proxy means test in developing
was as high as 46%. countries. A paper published by FAO (see Tiba 2011),
Developing countries encounter different problems of for example, suggests that means testing works best only
direct poverty assessment. A lot of people in these in settings where declared income is verifiable; but in
countries particularly those in poverty do not declare developing countries where income information is
their incomes and wealth to the authorities. This results notoriously difficult, proxy means testing becomes an
in inadequate data on poverty incomes and inaccurate alternative way to assess the wealth status of an
poverty lines. Professional workers; therefore, are individual or a household. According to a UN’s
unable to construct a direct poverty test or to base their publication (Ortiz 2007), proxy means testing is cheaper
decisions to provide or to withhold welfare benefit on than means tests and allows wider participation
clients’ financial status. Inapplicability of direct poverty opportunities for community groups and local
assessment also arises because social welfare in governments. One of the World Bank’s publications
developing countries is independent from taxation. As a (Grosh & Baker 1995) indicates that, in targeting the
consequence, use of income or assets data as criteria for poor, social workers gather many aspects of
providing social assistance benefit is not possible. information, thus, systematic use of collected data may
Literature has showed that direct poverty measurement improve targeting outcomes and increase the fairness
like means testing has been a tool for determining the and transparency of social programmes.
entitled amounts of social assistance benefit in most of Chile is known as the first country where the proxy
the welfare states for a long time. Arguments in favor of means test was originated (Kidd & Wyld 2011; Grosh &
the means test system are efficiency and adaptability. Baker 1995). The Chile model employed a form filled
Despite these advantages, means testing is not above out by a social worker to collect information on
criticism. Legros (2009) points out the complexity and household characteristics such as location, housing
unfairness of the means tested system in France. He also quality, household composition, and the works done by
mentions that the means tested benefit makes people the household members. Answers to these questions are
dependent on the system in the long-term. prescribed with scores which will help predict the
Stigmatization caused by means testing seems to be poverty or income level of each household. Other

Proceedings of Academics World International Conference, Copenhagen, Denmark, 23rd-24th October 2016, ISBN: 978-93-86083-34-0
4
Proxy Means Testing: An Alternative Method For Poverty Assessment

developing countries such as Bangladesh (Sharif 2009), judgmental sampling with a sample size of
the Philippines (Mapa & Albis 2013), Uganda approximately 3000 respondents in each population was
(Housson, Zeller, Alcaraz, Schwarze, & Johannsen applied.
2007), and Peru (Johannsen 2006), then, adopted the
proxy means testing as the method of eligibility 3.3. Data Collection
assessment. Social workers and welfare officers who worked with
the Social Development and Human Security Provincial
II. OBJECTIVES Offices around the county were responsible for
purposively choosing respondents in communities and
This article aims to show how an indirect method for social assistance clients to respond to the proxy means
poverty assessment or the so called proxy means testing tests. The study intended to draw representative samples
was developed. In conjunction with the first objective, from the two populations by trying to cover respondents
the article will suggest to welfare agencies that this from various socio-economic statuses in terms of
proxy means testing can be a valid testing method for gender, age, education, employment, income etc.
social assistance eligibility.
3.4. Data Analysis
III. METHODOLOGY 3.4.1. Validation of the Proxy Means Test The
validation took into account the capability to
The research was designed as a survey research which differentiate social assistance clients who were
collected data from people in communities and from presupposed to be persons in poverty from people in
persons who came for social assistance benefits communities who were believed to be more financially
throughout the country. The main aim of the survey was secure. The differentiation analysis compared the proxy
to test whether the indicators contained in the research means test data between the two sample groups. Only
instrument were valid enough to predict or to determine the indicator that showed a difference of at least 10% in
poverty levels. To achieve thid aim, the research terms of either an average value or a proportion between
launched the following procedure. the two samples would be kept as a question item in the
final version of the proxy means test.
3.1. Research Instrument Design
The research began with a review of literature 3.4.2. Proxy Means Test Criteria As the indirect poverty
concerning proxy means testing and group discussions assessment by the proxy means test must classify people
with social workers, welfare administrators, and welfare according to their financial security, the data analysis
recipients. The researcher expected that the literature needed to set up criteria for assessing people’s financial
review and group discussions would suggest a number status, To achieve this purpose, the analysis compared
of indicators that could predict a poverty level. After the the average proxy means test scores between the
literature review and group discussions, the researcher samples and related them to the poverty ratios among
decided to select 10 indicators for measuring poverty Thai people.
levels.
IV. RESULTS AND DISCUSSION
These 10 indicators consisted of educational levels,
personal illness, age, marital status, the burden of caring, 4.1. Validity of the Proxy Means Test Indicators
employment, debts, home ownership, housing All the 10 proxy means test indicators exhibited
conditions, and property owned. The researcher, then, apparent contrast between the two sample groups above
formed these 10 indicators into a proxy means test the predetermined 10% criterion set up prior to the data
which would become the instrument for the field survey. analysis. This means that each proxy means test
The proxy means test prescribed a score to each answer indicator is, to a large extent, able to differential social
of each question (see Box 1). The score was, actually, assistance clients from people in communities. In other
the number of the answer choice. If the respondent had words, the ten proxy means test indicators are capable of
no personal illness, for example, he/she would tick differentiating people in poverty from those who are
option ‘three’ under the question about ‘current personal more financially secure.
illness’ and would obtain a score of three. The total sum Consequently, we can identify people in poverty without
score of all items was a hint of the financial status of the paying attention to their incomes but by using the 10
respondent. proxy means test indicators to assess their socio-
economic status instead. In sum, people in poverty are
3.2. Populations and Sampling likely to have the 10 following characteristics in
The country-wide survey covered people residing in common:
communities and potential clients who came to the
Social Development and Human Security Provincial  Have a health problem
Offices for social assistance benefits during the time of  Are below 20 years of age or older than 60 years old
data collection. As the study was unable to construct a  Have a lower educational background
sampling frame, a non-probability sampling namely  Tend to be divorced or be a widow/ widower

Proceedings of Academics World International Conference, Copenhagen, Denmark, 23rd-24th October 2016, ISBN: 978-93-86083-34-0
5
Proxy Means Testing: An Alternative Method For Poverty Assessment
Box 1 The Proxy Means Test

Proceedings of Academics World International Conference, Copenhagen, Denmark, 23rd-24th October 2016, ISBN: 978-93-86083-34-0
6
Proxy Means Testing: An Alternative Method For Poverty Assessment

 Have insecure home ownership therefore, intended to introduce proxy means testing by
 Live in poor housing conditions which poverty levels are indirectly measured as an
 Own limited property alternative method for poverty assessment. By basing
 Have insecure employment status its initial analysis on literature and focus groups, the
 Have the burden of caring study developed the so called proxy means test which
 Are in debt consisted of 10 socio-economic indicators for
assessing people’s financial status. The study, then,
3.5. Proxy Means Test Criteria conducted a nation-wide survey to validate the proxy
Table 1 shows mean scores and standard deviations of means test. The findings indicated a number of
the social assistance client group, the people in conclusions. Indirect measurement by the proxy means
community group, and the whole sample group. The test comprising socio-economic indicators can be an
mean score of the social assistance clients is alternative to direct poverty measurement by income or
approximately 34.37 with a standard deviation of 10.93. assets testing. The ten indicators proved to be able to
This mean score is almost equal to the average value of measure poverty levels or to differentiate people in
the whole sample minus ½ of its standard deviation poverty from those with economic security comprises
(42.96-15/2). The people in community group who are employment status, health status, age, education,
most economically secure retains the highest average marital status, the burden of caring, debt burdens,
proxy means test score of 51.59 with the most dispersive property owned, home ownership, and housing
standard deviation of 13.50. conditions. The analysis of the national survey data in
If we take the proxy means test scores of the social relation to the World Bank statistics indicated that the
assistance client group (the mean score of 34.37 and its proxy means test scores from the social assistance
standard deviation of 10.93) as criteria for determining clients could be baselines for measuring poverty levels.
the national or whole country poverty levels, we will The value of 25 derived from the average proxy means
find a few interesting findings in relation to income test level scored by the social assistance clients minus
poverty in Thailand. First, the World Bank (2015) ½ of its standard deviation, for example, could be used
reports that 12.60% of the Thai population earned as the Thai national poverty threshold. This threshold
income below the national official poverty line. This can be compared with the national income poverty line
poverty ratio is almost equal to the figure of 13.6% of which is the conventional direct poverty measurement.
those who score lower than 25 in the whole sample Measurement of poverty by the proxy means test
group. The score of 25 is actually the average proxy shows more than economic status because it reveals
means test score in the social assistance client group socio-economic aspects of deprivation. Proxy means
minus ½ of its standard deviation (34.37-10.93/2). test scores, therefore, represent depth of poverty which
Therefore, the score of 25 can be comparable with the is useful for both social and economic development.
Thai official poverty line.
The same World Bank Report also mentions that the Apart from using the proxy means testing for indirect
income poverty headcount ratio at $4 a day (PPP) in assessment of poverty, this paper would suggest
Thailand is approximately 30.97%. This ratio is very another application of this indirect poverty
close to 32.2% of the whole sample who score lower measurement in judging social assistance eligibility.
than 34.97 which is the average proxy means test score From the above findings particularly that of the social
of the social assistance client group. As a consequence, assistance client group, the study can derive a proxy
the proxy means test score of 34.97 or approximately 35 means test scale for allocating general social assistance
can be comparable with the Thai poverty at the benefit. Box 2 shows how the scale functions to
international poverty line (PPP). determine an eligible amount of social assistance
benefit based on client’s proxy means test score.
Table 1. Average Proxy Means Test Scores and Their Standard Clients, whose proxy means test scores are in the Free
Deviations Area or below the threshold of 35, are entitled to a full
allowance. The score of 46 is the cut-out point at
which the provision of an allowance stops as persons
who score 46 or higher will be considered financially
secure, thus, ineligible for any amount of benefit.
Clients with scores from 36 to 46 are entitled to an
allowance determined either by a rigorous or a lenient
eligibility formula. The rigorous formula gives a
smaller amount of benefit than that computed by the
lenient eligibility one. The rigorous formula is
CONCLUSION AND RECOMMENDATIONS
appropriate for the situation where welfare agencies
have limited budgets or want clients to rely more on
In Thailand where information about incomes is
market incomes than on social assistance benefit.
inaccurate, the conventional methods of poverty
Welfare agencies may also choose to provide eligible
measurement such as poverty lines, an income test, and
clients with benefit larger
an assets test are rather impractical. This study,

Proceedings of Academics World International Conference, Copenhagen, Denmark, 23rd-24th October 2016, ISBN: 978-93-86083-34-0
7
Proxy Means Testing: An Alternative Method For Poverty Assessment
Box 2 Application of the Proxy Means Test

than nominal amounts by applying the lenient formula substituting these terms in either the rigorous or the
in determining social assistance allowances. lenient formula. If the current full allowance is 5,000
The welfare agency needs to declare a full or a baht/month (33 Thai baht = 1US$) and the client
maximum allowance (FA) before applying the proxy- scores 38 from the proxy means test, for example,
means-test formulas. The threshold (Th) and cut-out his/her rigorous and lenient allowances will be 4,000
(CS) scores are the static terms of 36 and 46. However, and 4,782 baht/month respectively.
the threshold and cut-out scores change corresponding
to national socio-economic circumstances. This REFERENCES
implies that the proxy means test scale and formulas
must be adjusted from time to time. [1]. Grosh, M., & Baker, J. (1995). Proxy Means Tests for
Targeting Social Programs: Simulations and Speculation,
A social worker obtains the client’s proxy-means-test LSMS working paper no.118. Washington D.C.: the World
score or the PMTS after interviewing him/her Bank.
according to the proxy-means-test questions. The [2]. Houssou, N., Zeller, M., Alcaraz, V.G., Schwarze, S., &
eligible amount of benefit (EA) is, then, derived by Johannsen, J. (2007). Proxy Means Tests for Targeting the
Poorest Households Applications to Uganda. Paper presented at

Proceedings of Academics World International Conference, Copenhagen, Denmark, 23rd-24th October 2016, ISBN: 978-93-86083-34-0
8
Proxy Means Testing: An Alternative Method For Poverty Assessment
the 106th seminar of the EAAE pro-poor development in low [9]. Pudney, S., Hancock, R., & Sutherland, H. (2006). Simulating
income countries: Food, agriculture, trade, and environment, the Reform of Means-tested Benefits with Endogenous Take-
Montpellier, France. up and Claim Costs. Oxford Bulletin of Economics and
[3]. Johannsen, J. (2006). Operational Poverty Targeting in Peru- Statistics, 68(2), 135-166.doi:10.1111/j.1468-0084.2006.
Proxy Mean Testing with Non-income Indicators. Brasilia 00156.x
(Brazil): International Poverty Centre (the United Nations [10]. Sharif, I.A. (2009). Building a Targeting System for
Development Programme). Bangladesh Based on Proxy Means Testing. Washington D.C:
[4]. Kidd, S.W., & Wylde, E. (2011). Targeting the Poorest: An the World Bank.
Assessment of the Proxy Means Test Methodology. Canberra: [11]. Tiba, Z. (2011). Targeting the Most Vulnerable: Implementing
Agency for International Development. Social Safety Nets. In A. Prakash (Ed.). Safeguarding Food
[5]. Legros, M. (2009). France: Minimum Income Schemes: From Security in Volatile Global Markets (509-528). Rome: Food
Crisis to Another, The French Experience of Means Tested and Agriculture Organization of the United Nations.
Benefits. Retrieved from file:///C:/Users /Get/ Downloads [12]. van Oorschot, W.J.H., & Schell, J. (1991). Means Testing in
/France_1_ 2009_EN%20(3).pdf Europe: A Growing Concern. In M. Adler, C. Bell, J. Clasen, &
[6]. Mapa, D.S., & Albis, M.L.F. (2013). New Proxy Means Test A. Sinfield (Eds.), The Sociology of Social Security (187-221).
(PMT) Models: Improving Targeting of the Poor for Social (Edinburgh education and society series). Edinburgh:
Protection. Paper presented at the 12th National Convention on Edinburgh University Press.
Statistics (NCS), Mandaluyong City, The Philippines. [13]. World Bank. (2016). Poverty and Equity Database. Retrieved
[7]. Moffitt, R. (1983). An Economic Model of Welfare Stigma. from http://databank.worldbank.org/data/reports.aspx?source =
American Economic Review, 73(5), 1023-1035. poverty-and-equity-database
[8]. Ortiz, I. (2007). Social Policy. New York: Department for
Economic and Social Affairs of the United Nation.



Proceedings of Academics World International Conference, Copenhagen, Denmark, 23rd-24th October 2016, ISBN: 978-93-86083-34-0
9